A Fast Inverse Warping Algorithm for Generating Novel View
生成当前视点目标图像的快速逆映射算法

By taking advantages of the features of epipolar lines and depth discontinuities in reference images, an efficient inverse warping algorithm for generating images of novel views is presented by combining multiple reference images taken from different viewpoints. The method proceeds in three steps. Firstly, the reference images are preprocessed for extracting edge pixels by their displacements. Secondly, a reference image is selected as the primary reference image to be warped inversely epipolar line by epipolar line to render the novel view. Because there exists correspondence between epipolar lines on different images and continuous segments formed by paris of edge pixels on epipolar lines are order-kept, the warping process is very fast by computing only pairs of edge pixels in the reference image to obtain generalized disparity of all pixels in the desired image. Finally, holes in the desired image are filled up through searching the corresponding pixels in other reference images. At this stage, some acceleration techniques are given by utilizing the edge information and occlusion relations between objects in reference images.